Built-in views

KubeStacks ships views for the projects most clusters run. When a cluster serves one of these kinds, its list gets the columns people look at first, its objects get a status that understands the project, related objects are a click away, and the changes its own CLI offers become actions. Like every change in KubeStacks, a view’s actions check your permissions first (you need patch on the kind), show the equivalent kubectl patch command, and are refused in a read-only cluster. Actions marked undo can be taken back from their notification.

​

cert-manager

cert-manager: certificates, their requests, and issuers. A certificate is Ready, Issuing, or critical with cert-manager’s reason. A certificate’s details add when it’s valid from, when it expires and renews, its lifetime and failed attempts. Expires reads like “in 30d”, and sorts by time.

​

Argo CD and Argo Rollouts

Argo CD applications, and Argo Rollouts. Applications take their status from Argo CD’s own health: Healthy, Progressing, Degraded, Missing or Suspended.

• Sync starts a sync, like the Sync button in Argo CD. It’s offered when no operation is running.
• Refresh asks Argo CD to compare the application with its source again.
• Promote moves a paused rollout past its pause. Resume unpauses one that was paused by hand.
• Restart replaces the rollout’s pods a few at a time, as its strategy says. It asks first.
• Abort stops a rollout in progress and sends traffic back to the stable version. It’s destructive, so it asks first.

Rollouts have the scale subresource, so you can also Scale them like a Deployment.

​

Flux

Flux: what it applies, and where it fetches it from. Statuses follow Flux’s conditions: Suspended, Stalled, Reconciling or Ready. Every Flux kind has the same three actions, like the flux CLI’s:

• Reconcile asks Flux to reconcile it now, like flux reconcile.
• Suspend stops Flux reconciling it (undo).
• Resume turns reconciliation back on, and asks for one right away (undo).

A Helm release that Flux manages links to its HelmRelease from Helm releases.

​

Gateway API

Gateway API: gateways and the routes attached to them. A gateway is Programmed once its controller has set it up. Routes are Accepted, or say why their gateway refused them, or why a backend can’t be resolved.

​

Karpenter

Karpenter: node pools, and the nodes it launched for them. A node pool’s details add its node class, CPU and memory limits, consolidation settings, and when its nodes expire.

​

KEDA

KEDA: event-driven autoscaling of workloads and jobs. A scaled object is Scaling or Idle, Paused, or critical with KEDA’s reason. Pause sets KEDA’s autoscaling.keda.sh/paused annotation, and Resume removes it.

​

External Secrets

External Secrets: secrets kept in sync with a vault or a cloud’s secret store. Sync now asks the operator to fetch the secret again right away.

​

Prometheus operator

The Prometheus operator’s kinds, as kube-prometheus-stack installs them. Prometheus and Alertmanager are Available, Degraded or Unavailable, as the operator reports them.

​

CloudNativePG

CloudNativePG: PostgreSQL clusters and their backups. Restart restarts the instances one at a time, replicas first and the primary last, with a switchover. It asks first.

​

Istio

Istio: traffic routing and security.

​

Velero

Velero: backups of cluster resources and volumes. Backups and restores are Completed, Partially failed, or critical with Velero’s phase.

​

Crossplane

Crossplane: the packages that extend it. A package is Healthy, Installing, or critical with the reason it isn’t installed or healthy.

​

CustomResourceDefinitions

Kubernetes’ own CRDs get a view too: their list shows each one’s Kind, Group, Scope and Versions, and their details add the plural, short names and stored versions.

​

Changing a built-in view

Your own view of a kind replaces KubeStacks’ for that kind. Copy the one you want to change from KubeStacks’ views into ~/.kubestacks/views, edit it, and press ⌘R (Ctrl+R on Windows and Linux). API resources then shows your file as the kind’s view.

​

Contributing a view

A view for a popular project is a welcome pull request. Add it to src/renderer/src/views, one file per project, with a comment linking to the project. Give its kinds the few columns people look at first, status rules when the conventions aren’t enough, and actions its own CLI offers (and that a patch can do). KubeStacks checks every view it ships when it starts, and the tests fail if one has a problem.